Help app to abandon the blind push of disturbing users, push release 2.0 product Smart push

A push now distributes 800 million messages a day to users, but they feel pushed too much, and the future push should be "pushing the right content to the right person at the right time and in the right scenario". Today, a push at the Gmic conference unveiled a push 2.0 Smart push for all developers in June, hoping to help app abandon the blind push that bothered users.

A push is a mobile (Android & IOS) messaging technology provider that services apps including Sina Weibo, where to go, China Merchants Bank, ink, singing, beauty Mito, etc. According to the introduction of the CEO Fang, a push is currently in the deployment of Third-party push services accounted for 90% of the market share, access to more than 80,000, so far the SDK access users accumulated more than 1.5 billion, if the heavy, the push of the service coverage of 500 million mobile customers.

At the time of last year, a push added an incremental update to the push and recommended the user with a possible application that might be preferred, under a message that prompts the user for a software update. A push 2.0 product Smart push is based on the recommendation that you might like to use it, hoping that in the future, users will be able to push the message through the experience without spamming.

How do you do it specifically? Return to the familiar word "big data". Fang Introduction, Smart Push can be based on user behavior data for user portrait, such as in the gender they can be "pure man, male sister, Pure girl, female, other" label, help developers in the follow-up based on the label to do accurate push, developers can customize the label. Developers can learn from the push process, feedback back data (such as opening rate, conversion rate, etc.) can help developers adjust the push strategy. In addition, in practice, developers can use a/b test for different attributes of the user group (based on the combination of tags) to do a push, so as to determine the different A/b variable conditions, how to improve the target users to copy, open the application even to the subsequent application of behavior (such as purchase) conversion rate. After a/b test, developers are able to push for high conversion rates for larger, more accurate people.

By the next quarter, the Smart push will continue to iterate, allowing developers to extract some of the samples for push, to find a part of the users who are particularly fond of pushing content. A push will automatically help developers find all users who are similar to these people based on their portrait, which can greatly reduce the user's harassment.

Based on proprietary tagging and testing by developers, Fang that the data can also create a mutual trust, effective data sharing and docking among developers. For example, if Mercedes-Benz to PPLive on the top of a video ads, the conversion rate will not be too high, because pplive many users are young people, can't afford Mercedes Benz. And the most accurate way to know if you can afford Mercedes is auto or Sina's car channel. This time auto can take this ad, through the corresponding data to the video ads accurate to pplive percent of users, because the conversion rate is greatly improved, even if each show is higher than the original price of several times, for Mercedes are attractive, and consumers can watch the experience of advertising feel better.

From the push of this conference, a push seems to have to work on the data on the meaning. Future push will also support developers in the application of the buried point, first to help developers to monitor whether the push can achieve the desired target. In addition, a push this year will also set up a special advisory department, hoping to learn from the company's large area of data to tell developers some experience.